https://github.com/microprediction/momentum

Running mean, variance, skew, and kurtosis
https://github.com/microprediction/momentum

kurtosis mean online-learning skewness variance

Last synced: 7 months ago
JSON representation

Running mean, variance, skew, and kurtosis

• Host: GitHub
• URL: https://github.com/microprediction/momentum
• Owner: microprediction
• License: mit
• Created: 2021-01-22T16:31:15.000Z (over 4 years ago)
• Default Branch: main
• Last Pushed: 2022-03-05T15:29:43.000Z (over 3 years ago)
• Last Synced: 2025-03-22T17:13:28.526Z (7 months ago)
• Topics: kurtosis, mean, online-learning, skewness, variance
• Language: Python
• Homepage: http://www.microprediction.org
• Size: 57.6 KB
• Stars: 11
• Watchers: 2
• Forks: 5
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# momentum ![tests](https://github.com/microprediction/momentum/workflows/tests/badge.svg) ![deploy](https://github.com/microprediction/momentum/workflows/deploy/badge.svg)

A trivial mini-package for computing the running univariate mean, variance, kurtosis and skew

- No dependencies ... not even numpy.

- No classes ... unless you want them.

- State is a dict, for trivial serialization. 

- Tested against scipy, creme, statistics

For multivariate covariance updating, maybe see [precise](https://github.com/microprediction/precise). 

### Install 

    pip install momentum

### Usage: running mean, var

    from momentum import var_init, var_update

    from pprint import pprint

    

    m = var_init()

    for x in [5,3,2.4,1.0,5.0]:

        m = var_update(m,x)

    pprint(m)

    

    

### Usage: running mean, var, kurtosis and skew 

    from momentum import kurtosis_init, kurtosis_update

    

    m = kurtosis_init()

    for x in [5,3,2.4,1.0,5.0]:

        m = kurtosis_update(m,x)

    pprint(m)

    

    

File an issue if you need more help using this. 

    

  

### Usage: running recency-weighted mean, var

    from momentum import rvar_init, rvar_update

    from pprint import pprint

    

    m = rvar_init(rho=0.01,n=15)

    for x in [5,3,2.4,1.0,5.0]:

        m = rvar_update(m,x)

    pprint(m)

    

This will switch from running variance to a weighted variance after 15 data points.